The Implementation of Ensemble Stacking Method on Convolutional Neural Network Model for Tomato Plant Disease Classification

Tomatoes are one of the horticultural commodities that are often used as food ingredients. Determining diseases in tomato plants can be started by examining infected plant parts such as leaves. However, examining tomato plant leaves requires much time and energy for farmers. Therefore, this study implements a machine-learning model that can classify tomato plant diseases based on leaf images. The research data uses a dataset of tomato plant leaf images obtained from the Kaggle repository consisting of 11 tomato plant disease classification classes. The classification model used is the result of implementing the Ensemble Stacking (ES) method on the Convolutional Neural Network (CNN) model with variations in testing the CNN_1, CNN_2, CNN_3, CNN_4, ES_CNN_1-2-3-4, ES_CNN_1-2-3, ES_CNN_1-2, ES_CNN_2-1, ES_CNN_1-4, and ES_CNN_1-2-4 models. These models were trained using 25,851 tomato plant leaf image data and evaluated using 1,100 data. The best model produced in this study was the ES_CNN_1-2 model which achieved an accuracy value of $95.18 \%$. This model was tested using original images taken with a camera and the results showed that the ES_CNN_1-2 model could accurately identify disease classification in tomato plants.